Date: Fri, 30 Oct 92 22:31:49 GMT
From: Christoph Kukulies <kuku AT acds DOT physik DOT rwth-aachen DOT de>
To: djgpp AT sun DOT soe DOT clarkson DOT edu
Subject: strange disk io problems

Hi all,
I'm having a very strange disk io problem for months now and everytime
I thought it would go away with the next djgpp release. I had it with 1.06
and now it occurs also with 1.08. I don't knwo if it's really a djgpp problem,
maybe it's just because djgpp (go32) is demanding more from motherboards,
hardware, disks, etc., but listen:

I have a 5 MB application which runs in 16 MB of physical memory. Maybe the
total memory usage is 10MB and despite of VGA memory accesses there is
very few paging involved, though disk i/o is often very heavy due to the
nature of the program - f2c is involved, too. Anyway, what happens is that
when the application is in graphics mode (800x600x256) that suddenly a 
black field appears in the upper left and the application locks.

It took a while until I found out that the black field was some text that
the os was writing out (black letters on black background) and that
this message was something like error on drive C:, sector not found,
Abort, Retry, Fail?

Strangely, this error did never occur on my development machine but
it occurs reproducable on other motherboard/disk combinations in my office.

Today I took my 340 MB Maxtor to one of the machines where the error occured
and replaced my disk against the 124 MB Maxtor 7120. BTW, all the failing
machines have Maxtor 7120.

Everything worked fine with my disk and the other motherboard. OK, I thought,
must be the disk. Got a disk from a local supplier (Quantum 245 MB - not
a factory fresh one - had some strange slow downs through formatting) and, voila
The error also appeared with this disk. 

Tomorrow I will do a test with a factory new 85 MB Conner. I also could observe
that the error message did occur much less often when the board was run
in non-turbo mode (Cache disabled).

I'm really asking myself if DJGPP does so weird things with the HW/SW
that election of the components becomes a significant issue?

Has anyone observed similar things? Is there eventually some test program
which does all these worst case conditions to the hardware/OS?

I wonder how it can happen, that such hard errors occur while when the machine
is used under Windows or DOS, everything seems to work smooth.


--Chris